@indoor-analytics/discrete-clustering

Clusters a bunch of paths by discretizing space with a given shape.

30 paths we want to cluster	hexagon clustering result

Corresponding graph	getClusteredPathFromGraph result

How to use

Add this line to ~/.npmrc to set up the package registry:

@indoor-analytics:registry=https://npm.pkg.github.com/indoor-analytics

In your project, install the package:

npm i --save @indoor-analytics/discrete-clustering

Import methods in your code:

import {clusterPaths, convertPolygonsToGraph, getClusteredPathFromGraph, Shape} from '@indoor-analytics/discrete-clustering';

Methods

Convenience method (recommended)

getClusteredPath calls all needed package methods to directly convert input paths into discrete-clustered output path.

function getClusteredPath (
    paths: FeatureCollection<LineString>,
    clusteringSettings: Partial<ClusteringSettings> = {}
): Feature<LineString>

You can tune clustering with some options:

Name	Default value	Description
targetDepth	5	Number of iterations wanted (the bigger, the smaller the discretization)
shape	Shape.Triangle	Shape used to discretize space

Manual clustering

If you need to tweak some settings or want to use a custom implementation, you can do the clustering process manually by calling all methods yourself.

// 1. Convert paths to cells
// 1.1 Recursive method (faster thus recommanded)
function clusterSpace (
    paths: FeatureCollection<LineString>,
    targetDepth: number,
    shouldColorCells = true,
    shape: Shape = Shape.Fit
): FeatureCollection<Polygon, {weight: number}>
// 1.2 Iterative method (slower)
function clusterPaths( paths: Feature<LineString>[], granularity: number, shape?: Shape ): FeatureCollection<Polygon, {weight: number}>;

// 2. Convert cells to graph
function convertPolygonsToGraph( cells: FeatureCollection<Polygon, {weight: number}> ): Graph;

// 3. Extract average path from graph
function getClusteredPathFromGraph( graph: Graph, start: string, end: string ): Feature<LineString>;

Discretization shape

Space can be discretized with different shapes:

• square
• triangle
• hexagon

Size of such shapes can be modified by playing with the granularity parameter of the clusterPaths method.